LIDAR Point Clouds Stitching Based On Dual Quaternion

LIDAR has such characteristics as high precision, strong autonomy, high automation and high efficiency. It can be widely used in many fields, such as navigation, aerial photomapping and geological disaster prediction. The spatial objects to be surveyed by LIDAR are complicated. In order to completely describe spatial objects, LIDAR point clouds of spatial objects are usually sampled from different perspectives, and then stitched by mosaic algorithm. The high precision LIDAR point clouds registration is the key to achieve overall point clouds data, in the meantime, to ensure the 3D topological reconstruction of spatial objects.Straight line is the high-grade geometric feature of LIDAR point clouds, and has strong geometric topology and constraint. In this paper, a new LIDAR point clouds stitching method based on dual quaternion is proposed. The conjugate lines between reference point cloud and point cloud to be stitched are expressed by plücker line, and the screw transformation of conjugate lines is described by dual quaternion. Then the collinear equation based on plücker line is established by geometric topology relation of the conjugate lines. Finally the relative position and attitude parameters between reference point cloud and point cloud to be stitched are solved by least square method.The reference LIDAR point clouds and the LIDAR point clouds to be stitched have the issue of different position, attitude and scale. A stitching method based on line cluster is proposed. Firstly, line cluster of the reference LIDAR point clouds and the LIDAR point clouds to be stitched is respectively established according to the topological relation of lines. Then conjugate lines of point clouds are represented in Plücker coordinate. Through the spiral zooming motion of line cluster of the LIDAR point clouds to be stitched, finally the scale of the two LIDAR point clouds is the same and the conjugate Plücker lines coincide with each other. The collinearity equations based on line cluster can solve the scale factor and the relative position and attitude simultaneously with high precision. The research provides data support for the complete expression of urban spatial objects.